Models › Qwen3 Coder 30B A3B (MoE) › RTX 3060 12GB
Can I run Qwen3 Coder 30B A3B (MoE) on RTX 3060 12GB?
With offload
Only with RAM offload. At Q4_K_M it needs about 20.8 GB, more than the 12 GB usable on the RTX 3060 12GB; with 32 GB system RAM, llama.cpp can split layers (MoE experts to RAM), at a speed cost. ESTIMATE
Verdict by quantization (8k context, f16 KV, 32 GB RAM for offload)
| Quant | Needs | RTX 3060 12GB (12 GB) |
|---|---|---|
| Q3_K_M | 17.4 GB | 🐢 Needs RAM offload |
| IQ4_XS | 18.4 GB | 🐢 Needs RAM offload |
| Q4_K_M | 20.8 GB | 🐢 Needs RAM offload |
| MXFP4 | 20.0 GB | 🐢 Needs RAM offload |
| Q5_K_M | 24.2 GB | 🐢 Needs RAM offload |
| Q6_K | 27.7 GB | 🐢 Needs RAM offload |
| Q8_0 | 35.5 GB | 🐢 Needs RAM offload |
| F16 | 64.9 GB | ❌ Doesn't fit |
Speed
No measured run for this pair yet, and we don't estimate MoE speeds. Submit a llama-bench run.
Memory figures are estimates from the model's parameter count and config.json (Hugging Face). How we calculate. Last verified with data from 2026-10-20.
FAQ
Can I run Qwen3 Coder 30B A3B (MoE) on the RTX 3060 12GB?
Only with RAM offload. At Q4_K_M it needs about 20.8 GB, more than the 12 GB usable on the RTX 3060 12GB; with 32 GB system RAM, llama.cpp can split layers (MoE experts to RAM), at a speed cost.
How much VRAM does Qwen3 Coder 30B A3B (MoE) need?
About 20.8 GB at Q4_K_M and 35.5 GB at Q8_0 with 8k context (estimate).
Qwen3 Coder 30B A3B (MoE) on other hardware: RTX 3090 · RTX 5090 · Apple M4 Pro (48 GB) · Apple M4 Max (128 GB)
Other models on RTX 3060 12GB: gpt-oss 20B (MoE) · Qwen3.6 27B · Qwen3.6 35B A3B (MoE) · Qwen3.8 27B · Gemma 4 31B · Gemma 4 26B A4B (MoE) · Gemma 4 12B · Qwen3.5 9B · GLM 4.7 Flash · Llama 3.1 8B Instruct
See all models for 12 GB VRAM.
Won't fit? Free hosted options: FreeAISwap.